one-password
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill is designed to manage high-value credentials. It interacts with the system clipboard using
pbpasteto ingest secrets and reads from~/.profileto identify service account tokens (MOLTY_OP_SERVICE_ACCOUNT_TOKEN). These operations are documented as core functionality for storing and retrieving 1Password items, and the skill provides clear instructions to never print these values to logs or chat. - [DYNAMIC_EXECUTION]: To securely perform operations like item creation, the skill generates temporary shell scripts in
/tmp. These scripts are created with restricted file permissions (chmod 700) to prevent local exposure and are deleted immediately after execution within atmuxsession. - [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill's metadata references the installation of the
1password-cliutility via Homebrew (brew), which is the official and trusted distribution method for this tool. - [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect injection as it ingests data from external sources like the system clipboard. However, it includes mitigation strategies such as basic validation of expected secret prefixes and strict boundary enforcement using a dedicated
tmuxsession for all operations.
Audit Metadata